Sumit Mohanty created AMBARI-10029:
--------------------------------------

             Summary: Node auto-recovery
                 Key: AMBARI-10029
                 URL: https://issues.apache.org/jira/browse/AMBARI-10029
             Project: Ambari
          Issue Type: New Feature
          Components: ambari-agent, ambari-server
    Affects Versions: 2.0.0
            Reporter: Sumit Mohanty
            Assignee: Sumit Mohanty
             Fix For: 2.1.0


Using blue-print, it is possible to perform a zero-touch install of hadoop 
clusters using Ambari. This is especially useful in the cloud environment. 
However, cloud environment also can be dynamic in the sense that nodes will get 
rebooted or reset to the original image.

Reset, being that the node (usually VM) gets reverted to original state where 
it joined the cluster. It is assumed that a reset node has ambari-agent 
installed and configured to communicate with the server. The node may also have 
all packages pre-instaled.

Node recovery is the feature to bring back a rebooted/reset online by starting 
or installing and then starting the host components that are already on the 
host.

In general, temporarily losing a node and then performing node recovery on a 
slave host should not affect the whole cluster. If its is a master node then 
there can be some disruption based on what is deployed on the master host and 
if HA is enabled for the master services or not.

Node recovery, discussed in this JIRA, only addresses the ability to 
automatically INSTALL/CONFIGURE/START host components on the node so that the 
desired state of the host component matches the actual state.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Reply via email to